Misconceptions in PMCC Hypothesis Testing

When I am teaching hypothesis testing with Pearson’s Product Moment Correlation Coefficient (PMCC), I find a few common issues keep cropping up. So let’s dissect some of the problematic areas that can cause this confusion.

🔙 Previous topic:

Review the master guide before tackling common PMCC errors.

1. “Correlation Means Causation” – Not Always

You calculate a correlation and get a high positive value. Then some of your pupils may exclaim: ‘A causes B!’ Sometimes I need to interrupt the lesson and ask a question: ‘What about a third variable?’

Take ice cream sales and the incidence of drowning. Both increase during the summer months. The correlations are strong, but one does not cause the other. I point out to my students that “PMCC simply reflects the strength and direction of a linear association and nothing more. It doesn’t reveal who’s calling the shots here!”

Have you ever noticed that it’s easy to get this mixed up? One of the common problems that may go unnoticed is moving directly from correlation to causation.

2. The sign of r is important

Another point that trips people up is interpreting the sign of r. Positive means both variables move in the same direction. Negative means they move in opposite directions. Simple, right? But here’s the kicker: the magnitude is often more important than the sign when you’re testing significance.

A quick classroom example: I write two examples of r-values on the blackboard. One with r = -0.85; another with r = 0.85. And of course, every time I do this lesson, the kids are shocked that the test requires the same for both.

“Hey, wait a minute; negative can be just as strong as positive? Yeah, for sure. First of all, strength counts; then direction.”

3. Mixing Up Critical Values And p-values

This mistake happens ALL the time. Sometimes students think that the key value is probability. NOPE. That’s not right at all. The key value comes from the chart using your degrees of freedom. But your p-value calculates the probability of seeing your data (or more) given that your null hypothesis is true.

Here’s a tip that I like to share with others: “The key value is the ‘hurdle’ and the test statistic is the ‘athlete.’ You want to compare the athlete to the hurdle.”

Let’s illustrate with an A-Level question. Suppose you have 8 pairs of data. Then degrees of freedom = n – 2 = 6. You refer to the tables to obtain the value of r for a 5% significance level. Afterwards, compare it to your value of r. If your r value happens to be higher, the null hypothesis goes for a toss.

4. Assuming That Linear Works

PMCC calculates only for linear correlations. This point can easily be missed. Often I ask my pupils to chart their data preliminary. Why? Because then one can easily see that perhaps a smiley face develops. And then PMCC can be close to zero although a strong relation may exist.

I remember one class where a student’s scatterplot clearly curved upward, but r came out near zero. “So, it’s like nothing’s happening?” they asked. I laughed. “Nope, PMCC just isn’t seeing the curve. You need something else—maybe Spearman’s rank correlation.”

It’s a gentle reminder: never trust numbers blindly. Always check the graph first.

5. Misunderstanding the Null Hypothesis

A good bit of the misunderstanding relates to what the null actually says. In PMCC, the null hypothesis often states that there’s no linear relationship between the population (ρ = 0). Sometimes it gets combined with ideas of random points, but it’s looking at the larger view.

I say to my students, “It’s like saying, ‘Let’s assume that nothing interesting happens to the population until we have evidence to the contrary.’ Then your test-statistic either provides you with enough evidence to question that or not.”

But yes, this is where careful interpretation matters here. r isn’t big despite being significant; it simply means that it’s improbable to have occurred by chance.

6. FORGETTING SAMPLE SIZE MATTERS

Here’s a subtle one: r = 0.4 might seem modest, but with 50 data points, it could be significant. With 5 points? Not likely. Sample size hugely affects the outcome of significance tests.

I always tell my pupils: “Suppose you toss a coin 3 times and get 3 heads. That’s strange? Not so. But toss a coin 100 times and get only 90 heads. Then it’s fishy.” That’s what we have here. The larger the sample size, the more accuracy we’ll have to spot a genuine relationship.

7. Two-Tailed and One-Tailed

Oh boy, the tail argument. One-tailed tests investigate a relationship in one way; two-tailed tests investigate in two ways. I’ve seen a number of pupils choose the incorrect tail simply because they ‘guess.’

Here’s my advice: always think about your research question first. You can ask yourself, “Am I only interested if it goes up, or could it go either way?” And it’s easy to forget that turning one way impacts your critical value a huge amount.

Quick Tips I Give Students In Class

Always plot your data first. Numbers are great, but graphs tell the story.

Review your assumptions. PMCC requires linear relationships, no outliers on either tail of the distribution, and interval/ratio scales of measurement. This is incredibly important.

Consider context. Statistical significance does not necessarily mean something is right.

Verify critical value. A small mistake can be quite common here.

Don’t forget to include the sign. This can give you important information since it indicates the direction of the relationship.

🧭 Next topic:

Revisit the fundamentals of PMCC hypothesis testing to avoid confusion.

Wrapping It Up

Honestly, hypothesis testing with PMCC isn’t complicated, but a person can easily get led astray. This requires careful thinking about what the actual implications of the numbers are as opposed to what one wants the results to mean. Always keep in mind that Correlation does not necessarily mean causation. One should never interpret the results without thinking of one’s assumptions.

During my classes, I like to tell my students: “Numbers never lie—but they never tell the entire truth.” A scatter plot can often help you avoid a lot of errors by simply visualising your ideas. A null hypothesis can also help.

So next time you tackle a PMCC problem, pause for a second. Plot the data, check the direction, look at the size, and ask yourself: “Does this really make sense in the real world?” Do that, and you’re already ahead of most students.

You can begin working on your revision with our live online A Level Maths sessions. Here you will better understand statistics, mechanics, and pure maths methods clearly. These 3 day courses provide a great way to meet topics like PMCC and to boost your understanding before exams.

About the Author

S. Mahandru is the Head of Mathematics at Exam.tips, specialising in A Level and GCSE Mathematics education. With over a decade of classroom and online teaching experience, he has helped thousands of students achieve top results through clear explanations, practical examples, and applied learning strategies.

Updated: October 2025

Misconceptions in PMCC Hypothesis Testing